[Spice-devel] [PATCH 06/11] worker: move some tree container functions

[Fabiano Fidêncio]

On Wed, Nov 11, 2015 at 5:10 PM, Frediano Ziglio <fziglio at redhat.com> wrote:
>>
>> On Wed, Nov 11, 2015 at 1:20 PM, Frediano Ziglio <fziglio at redhat.com> wrote:
>> > From: Marc-André Lureau <marcandre.lureau at gmail.com>
>> >
>> > ---
>> >  server/red_worker.c | 32 ++++----------------------------
>> >  server/tree.c       | 27 +++++++++++++++++++++++++++
>> >  server/tree.h       |  2 ++
>> >  3 files changed, 33 insertions(+), 28 deletions(-)
>> >
>> > diff --git a/server/red_worker.c b/server/red_worker.c
>> > index 5cdb348..e82317c 100644
>> > --- a/server/red_worker.c
>> > +++ b/server/red_worker.c
>> > @@ -933,30 +933,6 @@ static inline void remove_shadow(DrawItem *item)
>> >      free(shadow);
>> >  }
>> >
>> > -static inline void current_remove_container(DisplayChannel *display,
>> > Container *container)
>> > -{
>> > -    spice_assert(ring_is_empty(&container->items));
>> > -    ring_remove(&container->base.siblings_link);
>> > -    region_destroy(&container->base.rgn);
>> > -    free(container);
>> > -}
>> > -
>> > -static inline void container_cleanup(DisplayChannel *display, Container
>> > *container)
>> > -{
>> > -    while (container && container->items.next == container->items.prev) {
>> > -        Container *next = container->base.container;
>> > -        if (container->items.next != &container->items) {
>> > -            TreeItem *item = (TreeItem *)ring_get_head(&container->items);
>> > -            spice_assert(item);
>> > -            ring_remove(&item->siblings_link);
>> > -            ring_add_after(&item->siblings_link,
>> > &container->base.siblings_link);
>> > -            item->container = container->base.container;
>> > -        }
>> > -        current_remove_container(display, container);
>> > -        container = next;
>> > -    }
>> > -}
>> > -
>> >  static void display_stream_trace_add_drawable(DisplayChannel *display,
>> >  Drawable *item)
>> >  {
>> >      ItemTrace *trace;
>> > @@ -1034,7 +1010,7 @@ static inline void current_remove(DisplayChannel
>> > *display, TreeItem *item)
>> >                  continue;
>> >              }
>> >              ring_item = now->siblings_link.prev;
>> > -            current_remove_container(display, container);
>> > +            container_free(container);
>> >          }
>> >          if (now == item) {
>> >              return;
>> > @@ -2613,7 +2589,7 @@ static bool free_one_drawable(DisplayChannel
>> > *display, int force_glz_free)
>> >      container = drawable->tree_item.base.container;
>> >
>> >      current_remove_drawable(display, drawable);
>> > -    container_cleanup(display, container);
>> > +    container_cleanup(container);
>> >      return TRUE;
>> >  }
>> >
>> > @@ -3106,7 +3082,7 @@ static void red_update_area_till(DisplayChannel
>> > *display, const SpiceRect *area,
>> >          now->refs++;
>> >          container = now->tree_item.base.container;
>> >          current_remove_drawable(display, now);
>> > -        container_cleanup(display, container);
>> > +        container_cleanup(container);
>> >          /* red_draw_drawable may call red_update_area for the surfaces
>> >          'now' depends on. Notice,
>> >             that it is valid to call red_update_area in this case and not
>> >             red_update_area_till:
>> >             It is impossible that there was newer item then 'last' in one
>> >             of the surfaces
>> > @@ -3164,7 +3140,7 @@ static void red_update_area(DisplayChannel *display,
>> > const SpiceRect *area, int
>> >          now->refs++;
>> >          container = now->tree_item.base.container;
>> >          current_remove_drawable(display, now);
>> > -        container_cleanup(display, container);
>> > +        container_cleanup(container);
>> >          red_draw_drawable(display, now);
>> >          display_channel_drawable_unref(display, now);
>> >      } while (now != last);
>> > diff --git a/server/tree.c b/server/tree.c
>> > index bf50edf..ad31f09 100644
>> > --- a/server/tree.c
>> > +++ b/server/tree.c
>> > @@ -223,3 +223,30 @@ Container* container_new(DrawItem *item)
>> >
>> >      return container;
>> >  }
>> > +
>> > +void container_free(Container *container)
>> > +{
>>
>> Better add a check if container is valid as well.
>>
>> > +    spice_return_if_fail(ring_is_empty(&container->items));
>> > +
>> > +    ring_remove(&container->base.siblings_link);
>> > +    region_destroy(&container->base.rgn);
>> > +    free(container);
>> > +}
>> > +
>> > +void container_cleanup(Container *container)
>> > +{
>> > +    /* visit upward, removing containers */
>> > +    /* non-empty container get its element moving up ?? */
>> > +    while (container && container->items.next == container->items.prev) {
>> > +        Container *next = container->base.container;
>> > +        if (container->items.next != &container->items) {
>> > +            TreeItem *item = (TreeItem *)ring_get_head(&container->items);
>> > +            spice_assert(item);
>> > +            ring_remove(&item->siblings_link);
>> > +            ring_add_after(&item->siblings_link,
>> > &container->base.siblings_link);
>> > +            item->container = container->base.container;
>> > +        }
>> > +        container_free(container);
>> > +        container = next;
>> > +    }
>> > +}
>> > diff --git a/server/tree.h b/server/tree.h
>> > index 6249c28..01d4ff9 100644
>> > --- a/server/tree.h
>> > +++ b/server/tree.h
>> > @@ -82,5 +82,7 @@ static inline int is_opaque_item(TreeItem *item)
>> >  void       tree_item_dump                           (TreeItem *item);
>> >  Shadow*    shadow_new                               (DrawItem *item, const
>> >  SpicePoint *delta);
>> >  Container* container_new                            (DrawItem *item);
>> > +void       container_free                           (Container
>> > *container);
>> > +void       container_cleanup                        (Container
>> > *container);
>> >
>> >  #endif /* TREE_H_ */
>> > --
>> > 2.4.3
>> >
>> > _______________________________________________
>> > Spice-devel mailing list
>> > Spice-devel at lists.freedesktop.org
>> > http://lists.freedesktop.org/mailman/listinfo/spice-devel
>>
>> ACK!
>>
>
> The comment is a TODO for you or should I add the check?
> Personally this is just a move so I would not add anything.

I agree that, as it's just moving the code, is better to not change
anything on this patch.

>
> OT on a change: this check on the code
>
>    while (container && container->items.next == container->items.prev) {
>
> (the ==) looks odd. Is basically checking if container has 0 or 1 items.
> Looks like that more than cleaning is collapsing/optimizing the container.
>
> Frediano